Normalization / Compression workflow for H4n audio

Imaginate

Imaginate

Well-known member
I have audio recorded with an ME66 boom mic and the H4n. Its two people talking in a quiet room.

The levels recorded on the H4n tend to be quite low, I used a program called LEVELATOR to boost the levels and apply some compression, but I find the compression too extreme and I hear pumping...

I have access to soundtrackpro but haven't used it for treating my original h4n files... I'm curious as to settings some of you would use to process your audio files before the final mix.

I find normalization is a bit hit and miss if is one single peak in the waveform it will only the level as high as that one peak. I'm guessing a combination of subtle limiting and compression then normalization

yes no?
 
I encountered a similar situation with the Zoom H4N. I raised the levels in Final Cut Pro
using a FCP audio gain filter. With this you can boost the gain beyond the usual 12dB and
not have to resort to stacking audio tracks to sum the level higher. I then noticed a hum
in the audio. I sent the audio to SoundTrack Pro and using the analyzer found that there
was a power line hum which I had it remove. It was then quiet and up to a normal level.
 
Just say NO.

Mix your sound track. Any "auto" anything is going to lower your quality level, some a LOT.
Normalizing should not be used, stacking clips should not be used. In a picture editing situation, compression should not be used. You shouldn't be doing any of that unless it's in a quiet calibrated room. If for some strange reason you feel you must do your own sound then at least get a book or two and have a basic understanding of what is going on. STP is a better sound app than FCP but both are quite lacking in the tool/ workflow area.

BUT if you are not listening on calibrated monitors in a quiet space you should not be doing anything destructive (processing) to your audio. It would be like doing color correction on a monitor that hadn't been adjusted, a total waste of time.
 
Noiz2 said:
Just say NO.

Normalizing should not be used, stacking clips should not be used. In a picture editing situation, compression should not be used. You shouldn't be doing any of that unless it's in a quiet calibrated room. If for some strange reason you feel you must do your own sound then at least get a book or two and have a basic understanding of what is going on. STP is a better sound app than FCP but both are quite lacking in the tool/ workflow area.

BUT if you are not listening on calibrated monitors in a quiet space you should not be doing anything destructive (processing) to your audio. It would be like doing color correction on a monitor that hadn't been adjusted, a total waste of time.
Click to expand...

lets assume that we are in a quiet room and listening on calibrated monitors, how do we increase the gain on these waveforms to mix them properly? If you can't normalize or stack or compress or limit the only thing left to do is to lower the gain on all of the other sound sources to match the weak signals... that doesn't sound right.

Why is it such a strange thing to do your own sound? Most editors do.
 
For the reasons you've already listed. Location sound is very specialized. Of all the editors I know, there is only one who can deliver quality audio. The others send it to a audio post guy for the audio work.

Imaginate said:
Why is it such a strange thing to do your own sound? Most editors do.
Click to expand...
 
Imaginate said:
lets assume that we are in a quiet room and listening on calibrated monitors, how do we increase the gain on these waveforms to mix them properly? If you can't normalize or stack or compress or limit the only thing left to do is to lower the gain on all of the other sound sources to match the weak signals... that doesn't sound right.

Why is it such a strange thing to do your own sound? Most editors do.
Click to expand...


Since I'm not a sound guy, you can take my comments for whatever you want. When you ask the question 'normalize', or 'compress', it sounds like you want to do one or two operations to the entire sound signal. A 'normalize' operation, at least from what I've seen of the cheap 'audio tools', does this, 'find the 'peak' in a series of audio samples, set that to 'one' , then multiply every other sample such that they maintain the same ratio to the peak. What that does is make the peak 'bigger', and every thing else somewhat... but more importantly, it boosts the noise of the signal along with any other 'desired' signal.

In a word, there's no one filter that 'does it all'. The signal needs to be chunked up into segments that perhaps have different signal processing applied. This would not be at the 'final' stage, but at the beginning...

The signals I usualy deal with in processing aren't particularly amenable to being analyzed in a 'quiet room, with calibrated speakers'... unless of course the fillings in your teeth act as radio tuners...

But of the signal processing principles apply...
 
lets assume that we are in a quiet room and listening on calibrated monitors, how do we increase the gain on these waveforms to mix them properly? If you can't normalize or stack or compress or limit the only thing left to do is to lower the gain on all of the other sound sources to match the weak signals... that doesn't sound right.
Click to expand...

OK your not in a picture editing setup. You would, wait for it... Raise the gain, on the sounds that need it and lower the level on the sounds that need that. It's called editing (and or mixing - though usually that kind of rough work is done in the edit).

You don't "normalize because it sucks the natural dynamics out of the performances and adds noise where you don't need it. If you don't have control over the normalization level, many "normalizers don't, you gained everything up too much and now have to lower everything and you added noise in the process.

You don't "stack" because it's a stupid practice that can lead to phasing problems, makes mixes overly complicated and is totally unnecessary if you are using anything over a toy NLE (and the OP is on FCP so I know it has a gain filter).

Compression, doesn't raise the level so it really doesn't apply. Same with limiting. You could use either but you would still need to add gain to the low parts and isn't that what you were trying to do in the first place?



Why is it such a strange thing to do your own sound? Most editors do.
Click to expand...

Depends on what kind of editors your talking about. Home movie editors yes, wedding video and event videographers yes and no, - but they aren't really doing post sound. Very bare bones and slather some music over it. Not a knock that is the way most of it works. Ads, local, yes probably, but again not much in the way of post sound. National ads, I won't say never but not most of the time. Even when nationals are basically VO and slathered music they go to a post house to mix. Any film with a budget over $50,000, almost never. Any film over $100K very very rare and if so almost always just the "temp" sounds. TV shows, MOWs, any film with a real budget, NEVER.

IE, in any format where post sound matters good editors don't do sound. The exception would be out of monetary desperation.

Now most picture editors even on big budget films toss in some sounds and in fact one job of post sound is to "feed the edit" with sounds that fill in the holes. But other than as a reference and a very occasional grab (because the director fell in love with the sound in editorial) none of that sound is used in the final mix.

Picture editors as a rule make really bad sound editors, especially if they have anything to do with the picture edit. Sound editors are usually lousy picture editors also. It's not a deficiency, it's just that humans can only really focus on one or the other at any single time. Part of the reason that you need a sound person on set is that your cameraman can not physically focus on sound and picture at the same time. Even those that do both such as Walter Murch, don't do both on the same film.

There is always going to be someone who can do it all and do it well, but that is a one in a million shot.
 
Noiz2, I have to say that "Normalize" is a different process in different apps. The way I use Normalize (when I sometimes do) I use it in Steinberg WaveLab. All that does it set a level for the highest peak of the file. So if I set normalize to -2db, the loudest peak is at -2db, but its only done as a volume increase. No change in the dynamics. I think Soundflrge does something different.

And as for stacking: I wouldn't rule it out completely as long as you are still paying attention to the loudest peaks. I had a track of an old guy talking and I had my levels too low, so my audio track barely regestered, even turned up 12db. So I copied/pasted another instance of the audio, and another. There were a couple of peaks I had to tame using the point "P" command in FCP and my audio was fine.

I completely agree that you should NOT just do some automatic process. And if you need to remove noise, do it non destructively, or make a copy of the file to process and keep the original safe. iZotoprRX is a great tool, and the new version "II" has more editing features. The main thing is to play your whole project with an eye on the clip meters and adjust the peaks by hand. Normalizing, even the method I have, can be a problem if you are working with multiple clips because no 2 clips have the same peaks. So when you normalize them all to -2db, it takes a different amount on each clip to get to -2db, and you will have varying levels for each clip. Bring all clips up the same amount, then go adjust the peaks that get out of hand.

That's what I do when I don't want to mix the whole thing in Cubase.
 
Chadfish said:
Noiz2, I have to say that "Normalize" is a different process in different apps. The way I use Normalize (when I sometimes do) I use it in Steinberg WaveLab. All that does it set a level for the highest peak of the file. So if I set normalize to -2db, the loudest peak is at -2db, but its only done as a volume increase. No change in the dynamics. I think Soundflrge does something different.

And as for stacking: I wouldn't rule it out completely as long as you are still paying attention to the loudest peaks. I had a track of an old guy talking and I had my levels too low, so my audio track barely regestered, even turned up 12db. So I copied/pasted another instance of the audio, and another. There were a couple of peaks I had to tame using the point "P" command in FCP and my audio was fine.

I completely agree that you should NOT just do some automatic process. And if you need to remove noise, do it non destructively, or make a copy of the file to process and keep the original safe. iZotoprRX is a great tool, and the new version "II" has more editing features. The main thing is to play your whole project with an eye on the clip meters and adjust the peaks by hand. Normalizing, even the method I have, can be a problem if you are working with multiple clips because no 2 clips have the same peaks. So when you normalize them all to -2db, it takes a different amount on each clip to get to -2db, and you will have varying levels for each clip. Bring all clips up the same amount, then go adjust the peaks that get out of hand.

That's what I do when I don't want to mix the whole thing in Cubase.
Click to expand...

yes this is the problem that I have encountered with normalizing.

I'm more thinking along the lines of

http://www.izotope.com/products/audio/ozone/__maximizer.html

Ozone's proprietary IRC (Intelligent Release Control) technology lets you get the loudness you want without introducing pumping and other unwanted limiting side effects.

Using a product like this only in a way more subtler fashion to increase loudness but keeping the noise threshold on the recorded audio consistent.

Noise removal isn't really an issue I'm talking about. I just want consistent and even levels on my dialog and some very transparent limiting and compression when there is shouting or loud transients.
 
I'll take a dialogue track along with all the audio in a project, mix it in Cubase, and put a mellow compressor on it as I would for a vocal in a song. A limiter would do it in a way that does not increase the noise. A maximizer will pump up everything and act as a limiter to chop off the peaks. Maximisers are for mastering your final mix, not for pumping up a single track, though of course you can do that if you choose. I'd try to do as much as you can without processing the dialogue. I have Ozone, and it's got many tools, but I think I still prefer the sound I get from just mastering in WaveLab using a simple EQ, and Waves Ultramaximiser. Things seem to distort easily in ozone for me, and their maximiser is not forgiving like the Waves untramaximiser. If you must maximise, listen for distortion carefully.
 
Noiz2 said:
National ads, I won't say never but not most of the time. Even when nationals are basically VO and slathered music they go to a post house to mix. Any film with a budget over $50,000, almost never. Any film over $100K very very rare and if so almost always just the "temp" sounds. TV shows, MOWs, any film with a real budget, NEVER.
Click to expand...

I agree with those examples, but there is A LOT of video production happening outside of those examples where editors have to mix their own sound. We will always need the expert gurus in those areas to help guide us in technique and tools whether its mixing sound or color correcting our images, products like magic bullet looks wont make you a colorist but it can help you achieve a quick solution that is appealing to the client.

In the same way simple techniques or tools can help the editor be a better sound mixer.

The editor working on that wedding video, corporate video, etc, won't be sending out his audio to be mastered at Abbey Road but his audio is just as important to him. Why can't he have sound just as good? Having balanced audio levels isn't rocket science, so what are the methods and workflows?

Am I the only one who has imported wav files from the h4n or other recording devices that have been a little on the quiet side?
 
Actually, you are one of many who have to deal with low levels from the H4n. The answer is still the same. There is no magic button or plug-in called "fix my audio". It started with the source. If you are consistantly getting low levels at the recorder, then begin by adding more gain prior to the recorder by way of a quality pre-amp or mixer. Post audio will still require mixing, typically performed on a system which allows non-destructive automation where mixes can be auditioned prior to printing. This mixing includes checkerboarding the tracks to allow for careful use of eq automation to compensate for any proximity effects created by varying mic to source distance, tempering agressive peaks, and compensating for tonal changes resulting from mic/room interactions. How far you want to take it is up to you but be assured, there is a learning curve, and to get great sound requires much time spent focusing on the audio portion of post.
 
First; everyone in indie film seems to be hot on the H4n, but the H4n was aimed at musicians, so there are a few problems when the device is used for production sound. Musicians tend to work at relatively high db (decibel) levels; the sound sources are usually much louder than normal conversation/dialog levels, so the H4n and similar devices (like the DR-100) are biased to record at lower volume levels to compensate for the loud sound sources of musicians. So recording dialog on the set will always result in lower than normal dialog levels. And since music is louder than conversational dialog a little extra noise in the mic pres is not noticed until it is used for production sound where the noise from the pres becomes noticeable in audio post.

Chad's got it right - they are called different names by different companies, but there are Audio Suite plugs that will increase the overall gain of a clip. In Pro Tools there are two; Gain and Normalize. Normalize allows you to raise the highest peak up to 0db; gain allows you to raise the clip by percentage, a small but crucial difference as you can raise peaks to above 0db. The peaks above 0db distort like crazy but the rest of the clip is okay, so you just edit out the distortion and replace with the original audio - this is just a tiny piece of what dialog editing is all about.

Dialog mixing is another completely different trip. Every mixer I know has his/her own signal flow and choice of plug-ins. Just about the only things they seem to agree upon is that the noise reduction plug-in comes first, and volume levels are automated in the track. After that it's varied orders of compression, EQ, limiting, etc. I personally automate the hell out of the volumes - sometimes syllable by syllable - as I prefer not to use a limiter. Then my chain is noise reduction and EQ. The dialog receives very mild compression* on the stem buss to give the dialog little extra "pop".




* Usually 1.5:1, maybe up to 5:1 if it's an action scene to help the dialog cut through dense music and sound FX.
 
Chadfish said:
but I think I still prefer the sound I get from just mastering in WaveLab using a simple EQ, and Waves Ultramaximiser. Things seem to distort easily in ozone for me, and their maximiser is not forgiving like the Waves untramaximiser. If you must maximise, listen for distortion carefully.
Click to expand...

I think that is a good start some waves plugins for soundtrack pro.. the look ahead peak limiting is exactly what im looking for.
 
unclebob6958 said:
First; everyone in indie film seems to be hot on the H4n, but the H4n was aimed at musicians, so there are a few problems when the device is used for production sound. Musicians tend to work at relatively high db (decibel) levels; the sound sources are usually much louder than normal conversation/dialog levels, so the H4n and similar devices (like the DR-100) are biased to record at lower volume levels to compensate for the loud sound sources of musicians. So recording dialog on the set will always result in lower than normal dialog levels. And since music is louder than conversational dialog a little extra noise in the mic pres is not noticed until it is used for production sound where the noise from the pres becomes noticeable in audio post.

Chad's got it right - they are called different names by different companies, but there are Audio Suite plugs that will increase the overall gain of a clip. In Pro Tools there are two; Gain and Normalize. Normalize allows you to raise the highest peak up to 0db; gain allows you to raise the clip by percentage, a small but crucial difference as you can raise peaks to above 0db. The peaks above 0db distort like crazy but the rest of the clip is okay, so you just edit out the distortion and replace with the original audio - this is just a tiny piece of what dialog editing is all about.

Dialog mixing is another completely different trip. Every mixer I know has his/her own signal flow and choice of plug-ins. Just about the only things they seem to agree upon is that the noise reduction plug-in comes first, and volume levels are automated in the track. After that it's varied orders of compression, EQ, limiting, etc. I personally automate the hell out of the volumes - sometimes syllable by syllable - as I prefer not to use a limiter. Then my chain is noise reduction and EQ. The dialog receives very mild compression* on the stem buss to give the dialog little extra "pop".




* Usually 1.5:1, maybe up to 5:1 if it's an action scene to help the dialog cut through dense music and sound FX.
Click to expand...

I also have the H4N and it's indeed not usefull to record "silent sounds"... I use it to record ambiances sometimes and stuff...but some noise-reduction is needed mostly...and not always possible/

Combined with a good mixer it can be very usefull though
 
dre83 said:
Combined with a good mixer it can be very usefull though
Click to expand...

I found that unless you match clip levels between the mixer and the H4N you will very low levels.
Here's how I did this with an SD302 mixer. (Disclaimer, I do not own an H4N or the SD302. Therefore I'm not an expert on either one.
If any of this is wrong or incorrect hopefully someone on here will chime in.)

I don't believe the SD302 setup menu feature will allow you to increase gain enough to drive the H4N to clip
so an alternative is necessary.

I connected the SD302 mixer to the H4N with a XLR-to-1/4 inch jack. This maintains line level into the H4N I believe.
I then connected my FCP edit system audio out XLR to the SD302 input XLR.
In FCP, play the Bars and Tone in the Viewer. Then set the SD302 input gain to a level just before clipping (about +16dB).
Then on the H4N, set the record level to just before clip level (0bB). Now clip levels are matched.
From here on, you can ignore the H4N record levels and just watch the SD302 meter and avoid clipping.
 
Bruce, here's how to calibrate your mixer to your recorder/camera. On the 302, or the MixPre for that matter, turn on the tone generator. That sends a tone to your recorder. In the recorder, set the level of the incoming tone to -20db. Then you are virtually unclippable. You no longer need to change the levels on the recorder, only on the mixer. Now, with the H4N you may want to fudge the level a little lower because with consumer gear you may still get distortion. But you should get a nice full signal. There's no need to get FCP involved because you can't do that in the field.
 
I did try this approach, which is also documented in the user manual as the initial setup. However, this produced very low recorded levels. The next step is then to
use the advanced gain structure setup feature provided in the 302 setup menu. I did this over the phone with Sound Devices support. The 302 was not able to
drive the H4N to clip. We then went through what I described and my notes from that session are the gist of the posting. But once you have set it up in the office,
you don't need to change it out in the field. I admit it seems strange.
 
bruce7 said:
I found that unless you match clip levels between the mixer and the H4N you will very low levels.
Here's how I did this with an SD302 mixer. (Disclaimer, I do not own an H4N or the SD302. Therefore I'm not an expert on either one.
If any of this is wrong or incorrect hopefully someone on here will chime in.)

I don't believe the SD302 setup menu feature will allow you to increase gain enough to drive the H4N to clip
so an alternative is necessary.

I connected the SD302 mixer to the H4N with a XLR-to-1/4 inch jack. This maintains line level into the H4N I believe.
I then connected my FCP edit system audio out XLR to the SD302 input XLR.
In FCP, play the Bars and Tone in the Viewer. Then set the SD302 input gain to a level just before clipping (about +16dB).
Then on the H4N, set the record level to just before clip level (0bB). Now clip levels are matched.
From here on, you can ignore the H4N record levels and just watch the SD302 meter and avoid clipping.
Click to expand...

As Chadfish said...you need a tone... a tone gives you one signal at one level... Mostly (always actualy, but sometimes 440Hz is used for music) a 1kHz tone at -18dB for digital, -9dB for analog dBPPM

in VU = 0Vu at

@ +4dBu (this is the "electrical level")

Now for the H4N combined with a sound device mixer you have to attenuate the outputs... which is in the "secret menu" which I don't know by head at this moment....
 
You must log in or register to reply here.
Back
Top